Morphology Matters: A Multilingual Language Modeling Analysis

TL;DR

This study investigates how morphological complexity affects multilingual language modeling, finding that certain morphological features increase surprisal in models, but linguistically-motivated segmentation strategies can mitigate this effect.

Contribution

It provides a comprehensive analysis of morphological influences on language modeling using a larger, more diverse dataset and compares segmentation strategies for improved performance.

Findings

Morphological complexity correlates with higher surprisal in models.

Linguistically-motivated segmentation reduces the impact of morphology.

FST and Morfessor segmentation outperform BPE in handling morphology.

Abstract

Prior studies in multilingual language modeling (e.g., Cotterell et al., 2018; Mielke et al., 2019) disagree on whether or not inflectional morphology makes languages harder to model. We attempt to resolve the disagreement and extend those studies. We compile a larger corpus of 145 Bible translations in 92 languages and a larger number of typological features. We fill in missing typological data for several languages and consider corpus-based measures of morphological complexity in addition to expert-produced typological features. We find that several morphological measures are significantly associated with higher surprisal when LSTM models are trained with BPE-segmented data. We also investigate linguistically-motivated subword segmentation strategies like Morfessor and Finite-State Transducers (FSTs) and find that these segmentation strategies yield better performance and reduce the impact of a language's morphology on language modeling.